Have you had any experience creating a complete website using AI?

wait this is interesting quick q tho - when u say u genrated pages using python code that AI wrote, did u have to manauly tweak that code alot or did it just… work?? asking bc im trying to do smth similar and idk if i should just vibe code it or actually go the step by step route like u did ‍♂️
I actually created code blocks first like js & css files, header, footer, & post template, sitemap, etc. I have separately created a database file in csv format. I then wrote a python code that takes elements from database like headings, content, images, etc. and generates html pages. I had to refine it multiple times until I got the desired output though. But it's an easy process because I'm not writing code myself. Once the site is ready, I had to tweak things for improvements in design & performance.

If you want to create a site or saas, you should first fix core functionality and don't change it during the development. The design can be modified a thousand times without any issues as it's a harmless css file. If possible, write modular code, so that you don't need to change every page. For example, I separated header & footer from post content so that I can modify ads across all posts at once instead of updating hundreds of html files every single time. Such opportunities exist during development and you should keep an eye on them. You should also decide monetization in advance. If you code a php site yourself and then decide to add memberships, it becomes difficult later. Such features require complex code and you may need to upgrade from manual php code to a wordpress site. So, take these big decisions first and keep doing small updates over time.
 
Yes, had to keep correcting mistakes by ai , there were lots of bugs , although I used Claude and GitHub
 
I actually created code blocks first like js & css files, header, footer, & post template, sitemap, etc. I have separately created a database file in csv format. I then wrote a python code that takes elements from database like headings, content, images, etc. and generates html pages. I had to refine it multiple times until I got the desired output though. But it's an easy process because I'm not writing code myself. Once the site is ready, I had to tweak things for improvements in design & performance.

If you want to create a site or saas, you should first fix core functionality and don't change it during the development. The design can be modified a thousand times without any issues as it's a harmless css file. If possible, write modular code, so that you don't need to change every page. For example, I separated header & footer from post content so that I can modify ads across all posts at once instead of updating hundreds of html files every single time. Such opportunities exist during development and you should keep an eye on them. You should also decide monetization in advance. If you code a php site yourself and then decide to add memberships, it becomes difficult later. Such features require complex code and you may need to upgrade from manual php code to a wordpress site. So, take these big decisions first and keep doing small updates over time.
wait so u basically built ur own lil static site generator with python + csv database instead of just using wordpress from start?? thats kinda genius ngl. also quick q, when u say "fix core functionality first" do u mean like decide monetization/membership stuff even before writing any code at all, or just before u start scaling the site?? bc that part confuzed me a bit also how many times did u actually have to refine the python script b4 it gave decent html output lol, sounds like alot of trial n error
 
AI can save a lot of time when building a website. Has anyone here built a full site with AI?
 
wait so u basically built ur own lil static site generator with python + csv database instead of just using wordpress from start?? thats kinda genius ngl. also quick q, when u say "fix core functionality first" do u mean like decide monetization/membership stuff even before writing any code at all, or just before u start scaling the site?? bc that part confuzed me a bit also how many times did u actually have to refine the python script b4 it gave decent html output lol, sounds like alot of trial n error
You first need to create a project blue print and technical architecture documents before starting the coding. If you try to build a diesel-engine car and decide to make it into a electric car mid-way, it'll create a lot of confusion to you & the AI tool as well. First decide what's supposed to be a 100% truth during project development. It's absolutely fine if you want to launch a free product and then add monetization at a later stage but you must add it to project blue print so that the foundation is laid off considering all these constraints. It's okay to scale the project at a later stage but it should be adding, removing, replacing, or modifying the parts instead of redesigning the whole project. The core functionality should remain same throughout development and scaling.

I've refined the script 10-15 times before it worked. But it only took me 2-3 days because it's a small project. I've worked on mobile apps that took me 2-3 weeks. AI can't remember whole development, so it'll be difficult to develop projects with hundreds/thousands of lines of code. You need to keep correcting it or uploading files for reference. So, keep a changelog if you're building a huge project. For small projects, you can take decisions while you develop as they don't need much planning.
 
Hell Yeah and.. Hell No at the same time. If you are like me and have no idea what it takes it is better to start with a solid product and improve on it. It still takes a ton of time tuning up and I would suggest also getting a security specialist to take a look at the codebase. It definitely saves money on development costs and you can get the features you want the exact way you want it to perform. However it spits out what you put it. If your prompt is slop, the output will be slop too, especially in regards of content and UI.

Don't try to reinvent the wheel. Put your own tire on it.

One thing to note, If one ai gets stuck on a bug try a new chat or a different ai. For example; claude was stuck on a bug and would just go on a spree to find the bug and had not done any improvements on it. I got frustrated and went to codex. It did find out the issue within an hour and solved it within the next hour. Same thing happened in the codex and I went to claude to have it sorted. So going back and forth may help too.
 
Claude code is amazing. if you want to do Black hat things. Use evil open code.
 
Manus hands down is the best, in my experience. It works similar to claude code, but it also provides the hosting if you need it. You can also have it build for you to give you the code to host elsewhere.

What i usually do is start with a base LP that I spied and then prompt it with the changes that I want made. Its pretty quick and doesnt use up a lot of the tokens, I think even right now there may be a promotion where you can try it out at no charge (but dont quote me on that.)

I like that you are able to visually see the site inside manus wihtout having to move it to other service or open in browser on html. it makes editing things easier.
 
Manus hands down is the best, in my experience. It works similar to claude code, but it also provides the hosting if you need it. You can also have it build for you to give you the code to host elsewhere.

What i usually do is start with a base LP that I spied and then prompt it with the changes that I want made. Its pretty quick and doesnt use up a lot of the tokens, I think even right now there may be a promotion where you can try it out at no charge (but dont quote me on that.)

I like that you are able to visually see the site inside manus wihtout having to move it to other service or open in browser on html. it makes editing things easier.
interesting, ive been hearin alot of ppl mention manus lately but never actully tried it myself the fact it gives u hosting AND lets u export the code if u want is pretty clutch tbh, feels like it removes a lot of the usual friction

the workflow u described w/ spying a base LP then just prompting changes sounds super efficient too, way faster then buildin from scratch every time fr

question tho — how does it compare token/cost wise once u actully need to scale up or use it alot? like is it still cheap once ur past whatever free promo thing, or does it start addin up fast once u get into heavier projects? tryna figure out if its worth switchin over
 
the point about SEO being dead for beginners is what most people skip over. building the site is the easy part now, anyone can do it. i've spun up a couple full projects with claude code myself and the actual product was solid, but then you just sit there staring at zero traffic wondering what to do next.

honestly the abandoned transcription tool story above is the realistic outcome for like 90% of these. not because AI cant build it, it builds fine. its that everyone else is also building the exact same thing this month.

if you already have a traffic source or an audience then yeah AI dev is a huge time saver. if you dont, you're just making a nicer looking product nobody sees. thats the part id figure out first before touching any builder tool imo.
Of the abandoned tools and projects, most are because of the lack of consistent effort in researching competition, free marketing, outreach, open testing and feedback based upgrades. This has always been the case before but it is just more pronounced now with everyone capable of making a MVP using AI today. On the flipside, I guess it is much easier today to make tiny tools with AI because you need something very specific and it helps if you can code to ensure the codebase isn't fluffed up.

What I am trying to say is, if AI makes you something you are finding value in using, you would naturally open it for testing, take feedback (yours included too) and upgrade your tool till a point where competitor research suggests you have a niche opportunity. And if nothing else, you have a unique tool that works just right for you. Kind of like bash scripts and small codebases pre-AI that every developer had to help them along until they felt others would benefit a lot from it.
 
Back
Top